Machine learning reveals the most important psychological and social variables predicting the differential diagnosis of rheumatic and musculoskeletal diseases

摘要

There is an ongoing debate about the importance and the extent to which psychological and psychopathological factors, adverse childhood experiences, and socio-demographic characteristics are associated with the development of certain types of rheumatic disease. With the aim of contributing to knowledge on the subject, the present study uses machine learning modeling to determine the importance of 20 psychological and social variables in predicting two classes of rheumatic disease: inflammatory rheumatic and musculoskeletal diseases (RMD) (rheumatoid arthritis = RA, spondyloarthritis = SA, and Sjogren's syndrome = SS) versus non-inflammatory RMD, namely fibromyalgia = FM). A total of 165 French women with FM, RA, SA, and SS completed an inventory of personality traits, a psychopathology diagnosis questionnaire, and a fatigue/pain questionnaire. They also answered questions about adverse childhood experiences and socio-demographic characteristics. Random forest and logistic regression machine learning algorithms were used for data analysis. The main findings suggest that mistreatment during childhood ((MDA = 10.22), the agreeableness personality trait (MDA = 3.39), and somatic disorder (MDA = 3.25) are the main psychological and social predictors of the type of rheumatic disease diagnosed. The first two predictors (OR = 18.92 and OR = 6.11) are also more strongly associated with FM than with RA-SA-SS. Overall, adverse childhood experiences seem relatively more important than personality traits, psychopathological or demographic variables. The results of this study suggest that traumatic childhood experiences may lead to psychopathological disorders in adulthood, which in turn might underlie, at least in part, the development of FM. Since there are no imaging or biological markers of FM, the present findings contribute to the scientific literature offering information to help patients with FM understand their pathology. They may also provide physicians with more diagnostic information.